Let $\rho = K* \frac{P}{T} $ where K is s real constant number. So $\rho$ is related to $\frac{P}{T}$. Someone said that we can conclude from this:
$\frac{\Delta\rho}{\rho} = \frac{\Delta P}{P} - \frac{\Delta T}{T}$ Equation (I)
Delta is the common delta we use in physics or math. $f(x2) - f(x1) = \Delta f(x)$
My problem is how can we conclude Equation (I)? Is it true?
You can't because it's generally not true.
However, there is a similar relation between the differentials: the trick here is "logarithmic differentiation": for positive $z$, you have
$$ \mathrm{d} \log(z) = \frac{\mathrm{d}z}{z} $$
and logarithms turn products into sums. Here, it compues
$$ \mathrm{d} \log(\rho) = \mathrm{d} \log(KP/T) = \mathrm{d}\log(K) + \mathrm{d}\log(P) - \mathrm{d}\log(T) $$
Since $\mathrm{d}K = 0$, this equation leads to
$$ \frac{\mathrm{d}\rho}{\rho} = \frac{\mathrm{d}P}{P} - \frac{\mathrm{d}T}{T} $$
If you're not integrating the differentials to obtain the difference, then the things you can say about the differences $\Delta \rho$ and such are mainly through differential approximation: given a small variation $v$, you have $\Delta z \approx \nabla_v z$.
If you apply such a $v$ to the above equation and then substitute in the approximation, you get
$$ \frac{\Delta \rho}{\rho} \approx \frac{\Delta P}{P} - \frac{\Delta T}{T} $$
Note the approximation sign here; unlike the equation between differentials, this equation is generally not exactly true.